Is Windows on Arm still a 2017 thing?

[Jcmg62]

It's the 18th November. Black Friday is just days away. Christmas is around the corner.

Microsoft have no new product announcements due.

Is Windows on Arm no longer going to happen in 2017?

 

[Tien-Lin Chang]

According to the geekbanch result leaked, I really failed to see the point of W10A in the market right now. SD835 evens z7-8700 under x86 environment at best with similar TDP(~2W), the only good point provide by ARM SoC is the great integrity which means nothing to end user. LTE version of surface can give you the same internet connectivity and if you take a look at Dell yoga book, replace the strange touchpad with a normal design then throw in a bigger battery, power-saving FHD AMOLED screen, and a QCom X16. I don't think SD835 platform can have significant advantage even in battery life. I also failed to see ARM-based mobile device can benefit end user because they won't be cheaper than current choice you have on the market. So even W10A is in a stable stage now which I highly doubt I will don't think it will be a better choice when you buying a laptop/pad like device.

 

[Cruncher04]

According to the geekbanch result leaked, I really failed to see the point of W10A in the market right now. SD835 evens z7-8700 under x86 environment at best with similar TDP(~2W), the only good point provide by ARM SoC is the great integrity which means nothing to end user.

To put this into perspective, the Snapdragon 835 achieves about equal performance to Atom z7-8700 with emulated x86 apps - which in itself is an incredible achievement of the emulation technology. It is however more than 2x as fast with native code and more than 3x as fast for GPU workloads.

TDP is relatively meaningless, as most of the time the device is not running at the thermal boundaries and even when it is running at the thermal boundaries ARM CPUs do not need to throttle as much as x86 CPUs.

I don't think SD835 platform can have significant advantage even in battery life.

So you don't think that higher energy efficiency will results in longer battery life? That is very naive to say the least.

I also failed to see ARM-based mobile device can benefit end user because they won't be cheaper than current choice you have on the market. So even W10A is in a stable stage now which I highly doubt I will don't think it will be a better choice when you buying a laptop/pad like device.

They do not need to be cheaper if you just get the better device.
Having a much higher efficiency CPU inside is a big plus in my book. If you think about it, you can get either - a device with same form factor and performance at higher battery life or a lightweight small device with same battery life. There a lots of options if you just have a higher efficiency SoC.

 

[Tien-Lin Chang]

1) Talk about native performance of SD835 doesn't means much as it only happens with native UWP and there's tons of x86 app and packaged x86-UWP which SD835 have to run under emulator. That's why MS give it W10S as default and when people update it to W10 pro the shortcoming will be explored.

2) I don't think SD835 is more power effective under this emulating mode. First, the SD835 is meeting somewhere the lowest standard to run W10 smoothly, that's why other SD SoC are not adapting W10. So I don't think SD835 under W10 will stay as much time under optimized power-consumption curve as it under Android and once these SoCs are out of their optimized power-consumption curve they bleed out battery quickly. Second, with the two prototype released by ASUS and HP. They are even heavier than current x86 laptop/2in1 solutions and those extra weight I can only think of using in extra battery capacity. Simple math, the lenove yogabook is somewhere 650g with 32Wh battery and 13~15 hours using time, putting another 12000mAh power bank(~60Wh) together is still under 1kg while HP envy x2 is sitting at 1.2kg - do you really think atom will fall behind in using time?

3) with the LTE version of Android/Windows laptop/2in1 we already have almost every feature this SD platform stressed if not more. Without better cost-effective ratio to the customer why people would buy it?

 

[Cruncher04]

Talk about native performance of SD835 doesn't means much as it only happens with native UWP and there's tons of x86 app and packaged x86-UWP which SD835 have to run under emulator. That's why MS give it W10S as default and when people update it to W10 pro the shortcoming will be explored.

Wrong again. Windows on ARM device can run:
1) ARM 64 bit native UWP and Win32 apps
2) ARM 32 bit native UWP and Win32 apps
3) .Net/CLI apps

For all these app types no emulation is needed!

Only for
4) x86 32 bit native Win32 or Centennial UWP apps emulation is needed.

Simple math, the lenove yogabook is somewhere 650g with 32Wh battery and 13~15 hours using time, putting another 12000mAh power bank(~60Wh) together is still under 1kg while HP envy x2 is sitting at 1.2kg - do you really think atom will fall behind in using time?

HP envy X2 is around 670g without the cover (e.g. cover is relatively heavy) and 6.9mm thin (similar as Yogabook). However it has 2xCPU performance and 3xGPU performance compared to Yogabook - they play in a completely different league performance-wise. Also the yogabook only has a 10 inch screen with draws less power.
But even if we ignore the performance and screen size discrepancy for a moment:
Yogabook: Idle time with minimal display brightness, cellular/WLAN off: 1153 minutes. Browsing using WLAN: 498 minutes. And very typical for x86 devices is modern standby time of less than a week. (Reference: Notebookscheck.de)
"In our streaming video playback test, the Windows version of the Yoga Book ran for 8 hours on the dot." - cnet.com review

In summary a device with less than half the performance, smaller screen and about the same weight and thickness achieves much less battery duration than the HP envy X2!

Are serious or just trolling?

 

[Tien-Lin Chang]

According to xFastest, ASUS NovaGo only gives less than 13 hours of YouTube video playback through WiFi with 75% screen brightness. The actual battery life seems not as good as promoted.

Interesting

 

[justjun555]

According to xFastest, ASUS NovaGo only gives less than 13 hours of YouTube video playback through WiFi with 75% screen brightness. The actual battery life seems not as good as promoted.

Interesting

Thats not exeptional but good result.
I watched their video little but can't understand single word can you tell us more about it?

 

[Tien-Lin Chang]

Thats not exeptional but good result.
I watched their video little but can't understand single word can you tell us more about it?

Wow the video is a 50 minutes steam....I read the review article only.

Some points out of the review article:
1) no usb-c (ouch)
2) the machine they tested is actually a ES with 6 GB of ram instead of 4 GB in the retail model.
3) the UFS2.0 NAND performance is similar to SATA 3 SSD.
4) boot up time is around 36 second, not that fast and instant as promoted.
5) good video playback ability thanks to the SD835. 4K@60fps, 75Mbps video only take 40-50% CPU usage to play smoothly.
6) although an easy switch from win10s to win10 pro, the compatibility with x86 apps is still in doubt. Errors during installing test software happens so they are not sure how it affects the software even it seems runs and give results.
7) battery life, like I mentioned. Under the wifi video play setting it shows 12 hours 39 minutes when unplugged, 79% battery left after 2 hours 30 minutes playing and the estimated battery time is 9 hours. I think it might ends around 10-11 hours if they keep the test.
8) the current size and weight is the compromised results to save cost, it can be built lighter and smaller with ease.

Funny thing is even this product is already open for sale and test machines are giving to media, there's still not much touching to the actual daily use about the speed, stability and durability through all kinds of everyday apps.

I'm blaming the stupid pricing strategy from ASUS as well, the 699USD MSRP turns to be around 860USD retail price in Taiwan. If M$ can really bring MTK or even rockchip into the game I'm excited to see some 300-400USD product <- the price I would pay for the compatibility gamble.

 

[jlangner]

Good article on limitations:

https://www.thurrott.com/windows/wi...-finally-documents-limitations-windows-10-arm

 

[gold-stars]

MS drops the ball again looking at those.

So the Win32 emulator is not a 100% and Win32 apps that don't work will need to be recompiled to work with the emulator.

UWP is the only way to execute native Arm code and is 32-bit only no 64-bit support. Plus existing UWP Arm packages may not work either and also need to be recompiled to recognize the desktop Arm Win10.

x86 Win32 or UWP 64-bit apps are not supported.

What are the odds software on the Win Store gets updated to work properly with Arm Win10 ?, I'm not confident.

 

[Tien-Lin Chang]

Plus even before the first wave of SD835 devices hits the world market, the 845/850 rumor already everywhere and M$/QCom not even trying to stop it (release positive reviews of w10arm@SD835 if any?). Even thought the performance/battery life/software compatibility is not at the level people expect it to be, still something M$/QCom can/should do to keep this project hot/alive.


Just image customers decided to wait for 845/850 device and see how it goes then the selling of w10arm is terrible during 2018 (and early 2019 maybe) the M$ might cut loose this w10arm project again like they always did.....

 

[SymoClod]

Well it seems we are a little closer to finding out for ourselves as they have now announced the retailers who will sell them and the carriers that will offer the data packages. I guess we're not too far away from being able to buy one of these things.

 

[Jcmg62]

Well it seems we are a little closer to finding out for ourselves as they have now announced the retailers who will sell them and the carriers that will offer the data packages. I guess we're not too far away from being able to buy one of these things.

Microsoft need a win on this. Whatever variant of emulation they're using to run win32 on arm, it really needs to work well. If early adopters start turning up in hordes to complain about poor user experience it could sink Andromeda before it leaves the harbour

 

[justjun555]

Pricing is high lets see how is the actual experience compared to intel products.
Microsoft should target value for money products with Windows on arm instead of premium devices.

 

[grahamf]

https://www.engadget.com/2018/03/20/asus-novago-48-hours-preview/

They seem to be fine. They are not workstation level but they are good enough for most use cases.

 

[patwwh]

If CISC Type (e.g. Intel) CPU is PATA cable, RISC Type (e.g. ARM) is SATA cable. RISC Type is the way to go, in term of energy save, performance and future development.

In front of the Intel's license on x64-x86 emulator, Microsoft should not release 32 bits (support only 3.x GB Ram), but 64 bits Windows ARM version, and then announce that she will discontinue the support for 32 bits version in coming years, just like what Apple & Google have done. As I know, Intel hold license of 32bits while AMD hold license of 64bits. This announcement may bring certain pressure on Intel oppositely! Also, since Intel still have 32bits production lines, this announcement may politely force Intel to make concession on the license conflict of emulator.

 

[grahamf]

If CISC Type (e.g. Intel) CPU is PATA cable, RISC Type (e.g. ARM) is SATA cable. RISC Type is the way to go, in term of energy save, performance and future development.

In front of the Intel's license on x64-x86 emulator, Microsoft should not release 32 bits (support only 3.x GB Ram), but 64 bits Windows ARM version, and then announce that she will discontinue the support for 32 bits version in coming years, just like what Apple & Google have done. As I know, Intel hold license of 32bits while AMD hold license of 64bits. This announcement may bring certain pressure on Intel oppositely! Also, since Intel still have 32bits production lines, this announcement may politely force Intel to make concession on the license conflict of emulator.

Except the whole reason for this gong show is that you CANNOT just "force" Windows developers to migrate. They will either update their apps or they won't, and they most likely wont. That's why Windows RT was dead on arrival (well that and the windowing environment).


However, taking these baby steps means Microsoft can ease developers over to creating UWP apps that are hardware agnostic, then move to a future where they can slowly drop support for legacy apps and x86 and x64. S Mode is the first step, as it hints at a future where it's enabled by default on most computers (you can still disable it, but if you have difficulty doing so then you shouldn't be disabling it in the first place). Some time after that only business computers (and maybe gaming computers) will support legacy apps in any form.

 

[Tien-Lin Chang]

In summary a device with less than half the performance, smaller screen and about the same weight and thickness achieves much less battery duration than the HP envy X2!

Are serious or just trolling?

Not really

https://www.techspot.com/review/1599-windows-on-arm-performance/

With more reviews published, the real performance under the QCom hype is revealed.

The performance of SD835 platform is no match to N3450 under emulator and a justmatch with native apps. Z8700is about 90% of N3450 is terms of CPU performance. So what you claimed the lies than half performance is not true.

The battery life is not outstanding either. In different reviews the Asus Novato/HP X2 are sitting around 11-13 hours using time - looks like about 50% more than a proper atom platform with about 8 hours using time. When you consider the usual Atom platforms are rocking with ~30wAh battery while the SD835 platforms have ~50wAh battery I don't see how special that platform is in terms of power consumption. Not to mention the very poor comparability and stability with x86 apps.

Even the punch back from QCom against those reviews is soft like bum....

I think it's a wake up call for those blind believers?

 

[justjun555]

Not really

https://www.techspot.com/review/1599-windows-on-arm-performance/

With more reviews published, the real performance under the QCom hype is revealed.

The performance of SD835 platform is no match to N3450 under emulator and a justmatch with native apps. Z8700is about 90% of N3450 is terms of CPU performance. So what you claimed the lies than half performance is not true.

The battery life is not outstanding either. In different reviews the Asus Novato/HP X2 are sitting around 11-13 hours using time - looks like about 50% more than a proper atom platform with about 8 hours using time. When you consider the usual Atom platforms are rocking with ~30wAh battery while the SD835 platforms have ~50wAh battery I don't see how special that platform is in terms of power consumption. Not to mention the very poor comparability and stability with x86 apps.

Even the punch back from QCom against those reviews is soft like bum....

I think it's a wake up call for those blind believers?

We don't have choice either. Intel doesn't want to hurt their profit margins from core m and core i3 so they will never design quality atom chip.
Now it is upto Microsoft to improve x86 emulation and 64bt compatibility.

 

[Tien-Lin Chang]

We don't have choice either. Intel doesn't want to hurt their profit margins from core m and core i3 so they will never design quality atom chip.
Now it is upto Microsoft to improve x86 emulation and 64bt compatibility.

Atom actually improved a lot during its journey. I accidentally bricked my Z3736F tablet a while ago and it's doing ok under Windows 10 like the yoga book I've tried at the store. I only realized how much powerful it gained until I dig out my old netbook with Atom N280 and trying installing Windows 10 on it.....

So I'm kinda not getting it with the "never design quality atom chip" concept. It not losing to the best QCom can offer so if this is consider a poor work then I doubt what to expect from QCom.